INTRODUCTION

The homeodomain is a structural motif originally defined by homologous sequences found in the Drosophila homeotic selector proteins, a family of proteins responsible for specifying positional identity along the anterior-posterior axis of a developing embryo (1, 2). It is now well established that the homeodomain functions as a DNA binding domain and that homeotic proteins function as transcription factors that regulate the expression of genes necessary for segment identity (3). The homeodomain motif was subsequently found in numerous developmentally regulated proteins from species as diverse as sponges to mammals implying a functionally conserved mechanism of regulating gene expression during development (4). The extraordinary degree of conservation found among homeodomain sequences suggests that they evolved from a common ancestor through gene duplication. Presumably, sequence variations in the homeodomain and the surrounding protein accumulated as the duplicated genes acquired new developmental roles; however, sequence divergence must have also been restricted by molecular interactions necessary for function.

The mammalian Hox genes contain homeodomains that are more similar to the homeodomains found in Drosophila homeotic genes than any other homeodomain sequence. When expressed in fly embryos, some Hox genes can partially substitute for the function of their Drosophila orthologs (5, 6, 7, 8). This is an intriguing observation given that, with few exceptions, the only conserved sequence between orthologous pairs is the homeodomain itself and a short stretch of 6-9 amino acids just amino-terminal to the homeodomain. One explanation for this functional substitution is that the homeodomain, aside from recognizing specific binding sites in target promoters, also participates in interactions with other conserved proteins required for transcriptional regulation of effector genes (9, 10, 11, 12).

One group of highly conserved proteins required for basal levels of transcription in vitro are the general transcription factors (GTFs),¹ which form the transcription preinitiation complex. To establish high levels of expression, assembly of the GTFs is thought to be facilitated through protein-protein interactions with sequence specific DNA binding transcription factors (13, 14, 15). Interaction with the GTFs is usually mediated by a regulatory domain in the sequence specific transcription factor; however, these interactions may also involve the DNA binding domain (16, 17, 18). Of particular interest to this study is the recent demonstration that proteins containing either the even-skipped (eve) or Msx-1 homeodomains interact with an essential GTF, the TATA-binding protein (19, 20). Although the exact role of the eve homeodomain remains undefined, it is required mediate the interaction between TBP and the eve repression domain. The binding of the TBP to Msx-1 protein, however, occurs primarily through an interaction with the Msx-1 homeodomain.

Here we extend these reports of homeodomain-GTF interactions by demonstrating that the Abdominal-B (Abd-B) homeodomain can bind to the  subunit of TFIIE (TFIIE), but does not interact with TBP. Moreover, we show that while the diverged homeodomains Antennapedia and Abd-B interact with TFIIE, the even-skipped homeodomain does not. Our data indicate that the Abd-B homeodomain interaction with TFIIE occurs in vivo and that it can facilitate transcriptional activation in vitro. These data suggest that homeodomains mediate interactions with various basal transcription factors in a class-specific manner, which may have important implications for transcriptional regulation during development.

EXPERIMENTAL PROCEDURES

All transfection effector plasmids are based on pHY60-1 (41), which contains sequences encoding the Gal4 DNA binding domain (amino acids 1-147) cloned 3 of the Rous sarcoma virus long terminal repeat (RSV LTR). To make RSV-TFIIE, a XbaI-HindIII fragment from plasmid His6.IIE, encoding a histidine-tagged TFIIE (a gift from Dr. R. Roeder) was subcloned into pHY60-1 replacing the existing Gal4 DNA binding domain. RSV-NHD was made by replacing the Gal4 DNA binding domain of pHY60-1 with a DraI-XbaI fragment from plasmid pC18 encoding the entire DfdAC/Abd-B protein as well as some flanking Dfd cDNA sequences (25). The plasmid RSV-GN was made by PCR amplification of a DNA fragment encoding the N domain. The primers used for this reaction were GCTCTAGATACAATGGACACTATTCG and CGGGATCCCAGCGCCGTGTTCGT, which also introduced a BamHI and XbaI site for subsequent ligation into pHY60-1 forming an in frame fusion with the Gal4 DNA binding domain. Finally, plasmid RSV-GVP16 was made by substituting a XbaI-HindIII fragment from pGal4.VP16 (41) containing an in frame fusion of the VP16 acidic activation domain (amino acids 411-456) to the Gal4 DNA binding domain for a XbaI-BglII fragment spanning the Gal4 DNA binding domain in pHY60-1.

To make the reporter plasmid 33AdhCATHB, oligonucleotides containing a high affinity Abd-B binding site (42) were annealed, multimerized, and cloned into pBluescriptIISK⁺ (Stratagene) at the BamHI site. A plasmid containing five tandem binding sites was selected and a HindIII fragment from the 33AdhCAT effector plasmid (a gift from Dr. J. Lopez) containing the Adh TATA box and the CAT reporter gene was inserted 3 of the Abd-B binding sites. The reporter plasmid GB5CAT has been described previously in Yang and Evans (41).

DNA fragments encoding the three homeodomains and the N domain were amplified by PCR and cloned into pGEX-2T (43) using restriction sites introduced by the PCR primers. The sequence of each set of primers is available upon request. The induction and purification of the GST fusion proteins was performed as described in Ausubel et al. (44). The use of equal amounts of protein in the glutathione beads pull down assay was confirmed by equivalent staining intensities on Coomassie Blue-stained SDS-polyacrylamide gels. The TNT-coupled reticulocyte lysate system (Promega) was used to transcribe and translate the GTFs and TAFs in vitro under conditions recommended by the manufacturer. Plasmids encoding the Drosophila GTFs and TAFs were obtained from the laboratories of Drs. R. Tjian or J. T. Kadonaga, while the human TFIIE subunit clones were obtained from Dr. R. Roeder. Two micrograms of a preformed GST fusion protein-glutathione bead complex was incubated with the indicated amount of ³⁵S-labeled protein at room temperature (25 °C) for 30 min in a 60-µl reaction containing 1 × binding buffer (1 × BB). The 1 × BB contained 25 m Hepes (pH 7.8), 80 m KCl, 0.1% Nonidet P-40, 10% glycerol, 1.0 m DTT, 0.1 m EDTA, 2.5 µg/ml leupeptin, 1 mg/ml bovine serum albumin, and 5.0 m MgCl₂. The beads were pelleted, washed four times with 100 µl of 1 × BB, then boiled in SDS sample buffer (44). Bound proteins were resolved by 12% SDS-polyacrylamide gel electrophoresis for autoradiography. Band quantitation was performed using a Fuji Phoshor-Imaging system. To confirm biological activity of the GST-eve HD fusion protein, we performed a DNA binding experiment using the DNA insert from the plasmid p3102 containing known eve binding sites obtained from Dr. M. Biggin. Ten fmol of labeled DNA was incubated with 1 µg of either GST-eveHD fusion protein or GST protein prebound to glutathione beads in 150 m NaCl, 20 m Tris (pH 7.5), 0.25 m EDTA, 10% glycerol, 6.25 m MgCl₂, 0.05% Nonidet P-40, 1 m DTT, 50 µg per ml of sonicated herring sperm DNA, and 1 mg per ml bovine serum albumin at 25 °C for 30 min. The beads were washed twice with binding buffer without DTT, and bound DNA was fractionated on an agarose gel. Under these conditions, DNA binding to GST was not detected while the eve HD fusion protein bound approximately 50% of the input labeled DNA.²

The maintenance of the QT6 cells, transfections, and CAT assays were performed as described previously in Zhu and Kuziora (25).

A crude nuclear extract was prepared from HeLa cells (obtained from the National Cell Culture Center) as described in Dignam et al. (45). The in vitro transcription assay contained 300 ng of the 33AdhCATHB template, 30 ng of the GST-Abd-BHD fusion protein or 30 ng of GST protein and 100 µg of nuclear extract in a 50-µl reaction. The final concentrations of the other reaction components were: 12 m Hepes (pH 7.9), 12% glycerol, 60 m KCl, 0.12 m EDTA, 0.3 m DTT, 12 m MgCl₂, and nucleotide triphosphates at 600 µ each. The reaction was carried out at 30 °C for 1 h and then analyzed by primer extension (44). A primer complementary to the CAT reporter sequence (CATTGGGATATATCAACGGTGG) was used which produced an extended cDNA product 150 bases in length. For the antibody inhibition experiments, the nuclear extract was preincubated with the indicated amount of a polyclonal anti-TFIIE antibody (Santa Cruz Biotechnology) at 4 °C for 4 h before addition to the transcription reaction. Products of the primer extension reaction were fractionated on a denaturing polyacrylamide gel and quantitated using a Fuji Phosphor-Imaging system.

RESULTS

Transcription initiation requires a number of multisubunit GTFs, including TFIID, TFIIA, TFIIB, TFIIE, TFIIF, and TFIIH. In Drosophila, TFIID is composed of TBP and at least eight BP-ssociated actors or TAFs (21, 22). To investigate if homeodomains interact with GTFs or TAFs in vitro, a series of binding assays were performed with a GST-Abd-B homeodomain fusion protein and ³⁵S-labeled Drosophila TAFs_, including TAF30, TAF30, TAF40, TAF60, TAF80, TAF110_, and TAF150. We also tested the Drosophila GTFs, TBP, and TFIIB and human TFIIE and TFIIE. Of all GTFs and TAFs tested, binding at levels above that observed with GST alone was only observed with TFIIE. Approximately 50% of the input TFIIE bound to the GST-Abd-B homeodomain fusion protein (Fig. 1, lanes 1-4). The addition of 20 pmol of a double-stranded oligonucleotide encoding a high affinity Abd-B homeodomain binding site did not detectably alter the binding interaction with TFIIE or affect the negative interaction with the other GTFs and TAFs.²

Homeodomain interaction with TFIIE.

Homeodomains can be divided into classes based on sequence homology within and outside the homeodomain region (23, 24). Since the Abdominal-B homeodomain is representative of a relatively small class, we asked if sequence differences found in homeodomains of various classes can affect binding to TFIIE in vitro. A representative of the Antennapedia class, the Drosophila Antennapedia homeodomain (AntpHD), binds to approximately 60% of input labeled TFIIE (Fig. 1, lane 5). In contrast, a representative of a more diverged class of homeodomains, the Drosophila even-skipped homeodomain (eveHD), fails to bind TFIIE at levels above that retained by GST alone under identical conditions (Fig. 1, lane 6). The eveHD binds a labeled DNA fragment containing known eve binding sites, suggesting that the fusion protein retained biological activity.² These results indicate that sequence differences among homeodomain classes can affect TFIIE binding affinity in vitro.

We next asked if the interaction observed in vitro between the Abd-B homeodomain and TFIIE could occur in vivo. We reasoned that in cultured cells, artificially high concentrations of TFIIE might interfere with the ability of proteins containing the Abd-B homeodomain to regulate transcription by blocking interaction with the basal transcription complex. We have shown previously that the Abd-B homeodomain can weakly stimulate transcription, but at levels that would be insufficient for accurately measuring changes in response to high TFIIE concentration (25). To obtain higher levels of transcriptional activation, we used the DfdAC/Abd-B protein, which essentially consists of the Abd-B homeodomain fused to an activation domain found in the Deformed homeotic protein called the N domain (25). Sequences encoding this protein were cloned downstream of the RSV LTR promoter to create RSV-NHD (Fig. 2A). Transfection of RSV-NHD activates transcription approximately 20-fold above levels observed upon transfection of the 33AdhCATHB reporter plasmid alone (data not shown). To obtain a high cellular concentration of TFIIE, we constructed the expression plasmid RSV-TFIIE in which coding sequences from a TFIIE cDNA are placed downstream of a RSV LTR promoter (Fig. 2A). As shown in Fig. 2B, cotransfection of RSV-NHD with increasing amounts of RSV-TFIIE results in up to approximately 60% reduction in CAT activity without significantly affecting the level of transcription in the absence of RSV-NHD. We conclude that exogenous TFIIE interferes with the ability of the DfdAC/Abd-B protein to interact with the transcription complex.

The Abd-BHD interacts with TFIIE

Several lines of evidence suggest that the inhibition described above is due to an interaction between TFIIE and the Abd-B homeodomain rather than with other regions of the DfdAC/Abd-B protein or other transcription factors. First, a GST-N domain fusion protein does not bind to TFIIE (Fig. 1, lane 7). To confirm this observation in vivo, we cotransfected a construct encoding a Gal4 DNA binding domain-N domain fusion protein (RSV-GN; Fig. 2A) and RSV-TFIIE with a CAT reporter plasmid bearing 5 tandem Gal4 DNA binding sites (GB5CAT; Fig. 2A). As shown in Fig. 2B, the level of CAT reporter activation by RSV-GN is only reduced by approximately 10% in the presence of high concentrations of TFIIE, indicating that TFIIE does not significantly interact in vivo with the Gal4-N domain fusion protein. As a third control, we made use of the observation that the VP16 acidic activation domain interacts with TAF40, TBP, and TFIIB, but does not interact with TFIIE (26). As shown in Fig. 2B, cotransfection with RSV-TFIIE does not significantly affect the activity of a fusion protein consisting of the VP16 acidic activation domain and the Gal4 DNA binding domain. These experiments show that only a construct containing an Abd-B homeodomain is affected by overexpression of TFIIE, suggesting a specific interaction is possible in vivo between the homeodomain and the TFIIE  subunit.

We next asked if an interaction between the Abd-B homeodomain and TFIIE could enhance transcriptional regulation. We tested the ability of the GST-Abd-B homeodomain fusion protein to stimulate transcription in vitro using a HeLa cell nuclear extract. The GST-Abd-B homeodomain fusion protein stimulates transcription approximately 2.5-fold using the 33AdhCATHB reporter construct as a template (Fig. 3, columns 1-3). Activated transcription was not observed from the 33AdhCAT reporter construct which lacks Abd-B binding sites (data not shown). The modest activation mediated by the Abd-B homeodomain can be blocked by preincubating the nuclear extract with increasing amounts of anti-TFIIE antibody, without affecting basal transcription levels (Fig. 3, columns 4-11). These results support the hypothesis that the Abd-B homeodomain can contribute to transcriptional regulation through interaction with TFIIE.

can be blocked by anti-TFIIE antibody.

DISCUSSION

The three-dimensional structures of several homeodomain-DNA complexes reveal that homeodomains with diverged primary sequences are folded into remarkably similar globular structures (27, 28, 29, 30). Mutational analysis of protein-protein interactions involving the homeodomain have identified several amino acids that are involved in determining the specificity of the interaction (10, 31, 32). In general, these amino acids are considered to be solvent exposed as they are positioned on the opposite side of the homeodomain that contacts DNA. Residues in positions 11, 14, 15, and 22 of helix one, positions 32 and 36 of helix two, and position 24 in the intervening loop have been implicated in homeodomain-protein interactions and are identical or similar in the Antp and Abd-B homeodomains, but have different biochemical properties in the eve class homeodomains. One or more of these amino acids are likely to play a role in determining the specificity of the TFIIE interaction with the homeodomain.

How can an interaction between TFIIE and the homeodomain contribute to transcriptional regulation? At a biochemical level, TFIIE promotes the phosphorylation of the carboxyl-terminal domain (CTD) of RNA polymerase II by TFIIH (33, 34), but also inhibits a helicase activity shown by TFIIH that may be required to unwind the DNA prior to transcription initiation (35). Since the phosphorylated RNA polymerase II is the form of the enzyme that actively elongates transcripts (36, 37), the ability of a homeodomain to attract TFIIE to the initiation complex would serve to stimulate transcription by enhancing the kinase activity of TFIIH, resulting in a completely phosphorylated CTD. Alternatively, since TFIIE inhibits the helicase activity of TFIIH, it has been proposed that TFIIE might be removed from the complex following CTD phosphorylation (38). This is consistent with the observation that TFIIE binds unphosphorylated RNA polymerase II, but not to the phosphorylated form (39). Together with conformational changes that occur in the transcription complex during initiation that may alter access to the GTFs (40), the removal of TFIIE from the complex could be facilitated by interaction with a homeodomain. This scheme might have the added advantage of retaining TFIIE in the vicinity of the target promoter, thereby aiding subsequent rounds of initiation.

Unlike the results obtained with the Abd-B and Antp homeodomains, we failed to detect and interaction between the eve homeodomain and TFIIE. In contrast, an interaction between the eve protein and TBP has been reported (19). The eve homeodomain does not bind TBP in vitro, but it is required for the interaction between TBP and the eve repression domain and functions in some undetermined manner to stabilize the interaction. The mouse Msx-1 protein, which contains a homeodomain of the eve class, also interacts with TBP in vitro (20). In this case, a direct interaction between the Msx-1 homeodomain and TBP can be demonstrated. Of particular interest to this study was the finding that the Msx-1 protein does not interact with TFIIE (20). Thus homeodomains differ in their ability to interact with GTFs. Given the degree of sequence similarity within a homeodomain class, the interactions with specific GTFs may also be conserved. It follows that homeodomain-mediated interactions with other highly conserved proteins such as extradenticle/Pbx (10, 11) or SRF (12) might be retained among members of a homeodomain class. These interactions may be sufficient for orthologous vertebrate Hox proteins to partially mimic the transcriptional regulation of appropriate target promoters in fly embryos despite the lack of significant sequence homology outside of the homeodomain region.